International Journal of Machine Tools & Manufacture 45 (2005) 537–548

[Link]/locate/ijmactool

Expert spindle design system

Osamu Maeda, Yuzhong Cao, Yusuf Altintas*
Manufacturing Automation Laboratory, Department of Mechanical Engineering, University of British Columbia,
2324 Main Mall, Vancouver, BC, Canada V6T 1Z4
Received 27 July 2004; accepted 24 August 2004
Available online 5 October 2004

Abstract
This paper presents an expert spindle design system strategy which is based on the efficient utilization of past design experience, the laws
of machine design, dynamics and metal cutting mechanics. The configuration of the spindle is decided from the specifications of the
workpiece material, desired cutting conditions, and most common tools used on the machine tool. The spindle drive mechanism, drive motor,
bearing types, and spindle shaft dimensions are selected based on the target applications. The paper provides a set of fuzzy design rules,
which lead to an interactive and automatic design of spindle drive configurations. The structural dynamics of the spindle are automatically
optimized by distributing the bearings along the spindle shaft. The proposed strategy is to iteratively predict the Frequency Response
Function (FRF) of the spindle at the tool tip using the Finite Element Method (FEM) based on the Timoshenko beam theory. The predicted
FRF of the spindle is integrated to the chatter vibration stability law, which indicates whether the design would lead to chatter vibration free
cutting operation at the desired speed and depth of cut for different flutes of cutters. The arrangement of bearings is optimized using the
Sequential Quadratic Programming (SQP) method.
q 2004 Published by Elsevier Ltd.

Keywords: Spindle design; Expert system; Chatter vibration; Finite element method; Optimization

1. Introduction PRO-LOG, which is a programming language for expert

systems. The bearing arrangements are determined by the
The spindle is the main mechanical component in cutting operation type, and the required cutting force and
machining centers. The spindle shaft rotates at different life of bearings. Wong and Atkinson [2] demonstrated a
speeds and holds a cutter, which machines a material knowledge cell approach for diverse designs. They divided
attached to the machine tool table. The static and dynamic the knowledge cell into four parts; the Function, Selection,
stiffness of the spindle directly affect the machining Graphics, and Logic cells.
productivity and finish quality of the workpieces. The For design optimization of spindles, Yang [3] conducted
structural properties of the spindle depend on the dimen- static stiffness to optimize a bearing span using two bearings,
sions of the shaft, motor, tool holder, bearings, and the and described the methods used to solve the multi-bearing
design configuration of the overall spindle assembly. spans’ optimization method. Taylor et al. [4] developed a
This research considers spindle component selection and program which optimizes the spindle shaft diameters to
configuration using the proposed expert system based on the minimize the static deflection with a constrained shaft mass.
digital knowledge base. The expert system with fuzzy logic The Downhill Simplex Method is used to find the optimum
is implemented as the selection system. value. Lee and Choi [5] conducted an optimization design in
Eskicioglu et al. [1] developed a rule-based algorithm which they minimized the weight of the rotor-bearing system
for the selection of spindle bearing arrangement using with the augmented Lagrange multiplier method. Chen et al.
[6] and Nataraj and Ashrafiuon [7] demonstrated the
optimization results to minimize the forces transmitted by
* Corresponding author. Tel.: C1 604 822 5622; fax: C1 604 822 2403. the bearings to the supports. Wang and Chang [8] simulated a
E-mail addresses: altintas@[Link] (Y. Altintas), [Link]
[Link]/~mal (Y. Altintas).
spindle-bearing system with a finite element model and

0890-6955/$ - see front matter q 2004 Published by Elsevier Ltd.

doi:10.1016/[Link].2004.08.021
538 O. Maeda et al. / International Journal of Machine Tools & Manufacture 45 (2005) 537–548

depending on the machining application. In addition,

most of them optimize design parameters, such as shaft
diameter, bearing span, and bearing preload, to minimize
the static deflection. This paper considers more than two
bearings in the spindle model and takes into account the
chatter stability that is totally related to the dynamic
properties of the spindle.
The overall expert spindle design system is outlined in
Fig. 1. The design of the spindle with optimized bearing
spacing is automated using the requirements set by the
machining application, expert spindle design rules, cutting
mechanics, structural dynamics and chatter stability of
milling process.

2. Expert system for spindle design

The expert system for spindle design is introduced here

to facilitate the design process using past experience and
knowledge.
The expert system combined with the fuzzy logic is
used as the selection system of components for the spindle
design in order to handle uncertainties in the design
process as illustrated in Fig. 2. The required input data for
the spindle design, such as the cutting torque and power,
are computed using the laws of cutting mechanics, as
described in [11]. The input data is entered into the fuzzy
inference system, which is established by design experts,
Fig. 1. Outline of expert spindle design system. and is fuzzified using membership functions. The
Mamdani method is used as the inference system. The
compared it to the experimental results. They concluded that fuzzified values are applied to the fuzzy rules and
the optimum bearing spacing for static stiffness does not aggregated using the maximum method. The result of
guarantee an optimum system dynamic stiffness of the the aggregation is defuzzified using the centroid method,
spindle. Hagiu and Gafiranu [9] demonstrated a system in and a defuzzified number is obtained. The simple
which the bearing preload of the grinding machine is defuzzified number is applied to the selection rule for
optimized. Kang et al. [10] conducted static and dynamic the spindle components. An external database, which
spindle analysis by using an off the shelve FE system with an includes material cutting coefficients, is connected to the
added rigid disk and non-linear bearing model. fuzzy inference system, which users can access. The
The previous research used only two support bearings, supervising engineer, who is permitted to maintain this
although practical spindles may use more bearings expert system, can modify the membership function

Fig. 2. Internal structure of expert system for spindle design.

 O. Maeda et al. / International Journal of Machine Tools & Manufacture 45 (2005) 537–548 539

Table 1
Properties of transmission type

and database when the tendency of the fuzzy terms, such 2.1. Selection of transmission type
as ‘high’, ‘middle’, and ‘low’ changes as the technology
evolves. In this article, transmission and lubrication Torque has to be transmitted from the motor to the spindle
types are determined using the expert system with fuzzy shaft. There are a number of transmission types [12] and the
logic. main design configurations are gear (G), belt-pulley (B),

Fig. 3. Process of transmission type selection.

540 O. Maeda et al. / International Journal of Machine Tools & Manufacture 45 (2005) 537–548

one tooth period

Tmax c Z maxðTc Þ (4)

The cutting power Pt is found from,

X
N
Pt Z V Ftj ðfj Þ; with fst % fj % fex (5)
jZ1

where VZpDn is the cutting speed and n is the spindle

speed. Typically, the worst cutting condition, the slot
milling (fstZ0, fexZp) is considered in sizing the spindle.
Step 2. Spindle motor specifications. The spindle motor
specifications must be determined to identify the trans-
Fig. 4. Power and torque diagram.
mission type. The power and torque diagram is shown in
direct-coupling (D), and motorized (M) types as summarized Fig. 4. The relation between motor power Pmo and motor
in Table 1. torque Tmo can be expressed as follows
Fig. 3 illustrates the process used to determine the type of 2p
transmission between spindle shaft and spindle motor. Pmo Z n T (6)
60 mo mo
The process of transmission type selection can be
described as follows: where nmo is the motor rotation speed in rpm. nmax, as shown
Step 1. Evaluation of required cutting torque and power. in Fig. 4, is the maximum motor speed.
The required cutting torque is evaluated from the given Step 3. Classification. The spindle motor specifications
cutting conditions, and compared against the spindle motor are checked by applying the maximum cutting torque Tmax c
specifications. The instantaneous cutting torque Tc is and the required corresponding spindle speed n in Fig. 4.
evaluated based on the laws of cutting mechanics as The expert system checks if the maximum cutting torque is
described in [11] below the motor torque Tmo. The system also checks
whether the maximum speed nmax of the motor is greater
DX N
than the spindle speed used in the target machining
Tc Z F ðf Þ; with fst % fj % fex (1)
2 jZ1 tj j application. T ac and nacmax are the actual spindle
torque and actual maximum spindle speed, respectively,
where D is the diameter of the milling cutter, and N is the and are expressed as follows
number of teeth of the cutter, fj is the instantaneous angle of
Gs Gmo
immersion, and fst and fex are entry and exit angles for the Tac Z Tmo ; nacmax Z nmax (7)
cutter. The tangential cutting force Ftj is given by; Gmo Gs

Ftj ðfj Þ Z Ktc ahðfj Þ C Kte a (2) where Gmo and Gs are the gear sizes of the motor side and
spindle side, respectively.
where a is the axial depth of cut, Ktc and Kte are the cutting If the required speed and torque for the target
and the edge force coefficients, respectively. The material- machining application are higher than the capacity of
dependent cutting coefficients are evaluated from cutting the motor, the designers are warned to make changes in
experiments and stored in a database. The chip thickness the design. Table 2 shows the classification for the
variation (h(f)j) is expressed as follows transmission selection.
hðfj Þ Z c sin fj (3) Step 4. Application of fuzzy logic. In the classification
example shown in Step 3, the transmission possibilities
where c is the feed rate (mm/rev-tooth). are reduced from the gear, belt-pulley, direct
The cutting torque Tmax c required for the spindle motor coupling, and motorized type, to just ‘gear/belt-pulley’
is the maximum value among the instantaneous torque Tc in or ‘direct-coupling/motorized’ mechanism. The fuzzy

Table 2
Classification for transmission selection

Gear Belt-pulley Direct coupling Motorized Not available

(1) Tmax cOTmo and nOnmax (GsZGmo) #
(2) Tmax cOTmo and/or nOnacmax (GssGmo) #
(3) Tmax c%Tmo and n%nacmax (GssGmo) (3-1) Tmax cO15,000 #
(3-2) Tmax c%15,000 # #
(4) Tmax c%Tmo and n%nmax (GsZGmo) # #
 O. Maeda et al. / International Journal of Machine Tools & Manufacture 45 (2005) 537–548 541

Fig. 5. Membership function of torque. Fig. 7. Membership function of low cost vs. high accuracy.

logic process for the transmission selection between the 1 and 10, which indicates the conflict between the cost
gear and belt-pulley types are shown as follows. and accuracy.
Gear versus belt-pulley. The transmission type must also
be fuzzified. The membership functions of the gear or
2.2. Selection between gear type and belt-pulley types belt-pulley are shown in Fig. 8. These membership
functions are used in the implication process.
The following fuzzy sets are used to select the The Fuzzy rules shown in Table 3 are applied to the
transmission between gear and belt-pulley types. These fuzzified values of torque, speed, cost versus
functions are set by referring to bearing catalogs [13–15] accuracy, and gear/belt transmission type via
and the lead author’s 5 years of experience in a Japanese membership functions. These rules are defined from
machine tool company as a spindle designer. design principles.
Torque. The belt-pulley type cannot transmit large torque The following example illustrates the expert spindle
compared to the gear. The membership function of the design procedure. The required cutting torque, maximum
torque set is shown in Fig. 5. The torque can be classified as spindle speed, and the weight value for cost/accuracy are
‘Small’, ‘Middle’, and ‘Large’. given as 180 N m, 6000 rpm, and 7, respectively. These
Spindle speed. The belt may expand due to the three numbers are applied to membership functions. Note
centrifugal force at a high speed range, which prevents the that only membership functions for low and middle
transmission of large torque. Fig. 6 shows the membership speed are used for the selection of gear or belt-pulley
functions of the spindle speed. The belt and gear type type. All seven rules are applied to the membership
transmissions are usually not available at a spindle speed functions. Fig. 9 shows the implication process for rule
range higher than 15,000 rpm. As a result, the membership 4. Torque, spindle speed, and cost/accuracy are fuzzified
functions of ‘Low’, and ‘Middle’ are used when the gear or as 0.133, 0.666, and 0.3, respectively. Since the
belt-pulley type transmission is selected. minimum operator is used for implication, the minimum
Low cost versus high accuracy. The belt-pulley fuzzified number 0.133 is applied to the membership
transmission parts can be manufactured at a low cost function for gear/belt-pulley type.
because of their simplicity. In contrast, the gears are Fig. 10 shows the whole process of fuzzy logic. After
more costly due to the design and manufacturing implicating all seven rules, seven membership functions are
complexity involved. On the other hand, due to the obtained as shown on the right hand side of Fig. 10. These
belt tension applied at the pulley point, the spindle shaft seven membership functions are aggregated and a final
deflects and the rotation accuracy of its spindle is not as membership function is obtained, as shown in the lower
good as the gear type. The membership functions of low right of Fig. 10. From the final membership function, the
cost or high accuracy sets are shown in Fig. 7. The users centroid point is computed and the value of the lateral axis is
have to select an integer weighting number between taken as a defuzzified number.

Fig. 6. Membership function of spindle speed. Fig. 8. Membership function of gear vs. belt-pulley.
542 O. Maeda et al. / International Journal of Machine Tools & Manufacture 45 (2005) 537–548

Table 3
Fuzzy rules for transmission selection (gear/belt-pulley)

The final defuzzified value GB becomes ‘4.19’, as shown vibration is an important issue for machine tools since it
in Fig. 10, and the transmission type is determined with the may lead to spindle, cutter and part damages.
following selection rule: There are significant number of parameters in a typical
spindle design process, such as the dimensions of the spindle
IF GBR5 THEN transmission is belt-pulley type shaft, housing, and collars. However, the most effective
ELSE IF GBR5 THEN transmission is gear type design parameters need to be selected to optimize the spindle
design in practice. There are numerous constraints on the
Therefore, the gear type is chosen for the transmission geometric design of spindle parts, and design dimensions
since GB is equal to 4.19. which are usually coupled with each other. For example, if
the diameter of the spindle shaft changes, the bore diameter
of the housing also has to be changed, where more parameters
3. Optimization of bearing locations need to be taken into account which may lead to a
convergence problem in optimization algorithm. Since the
In order to apply the optimization to the spindle design, objective function is highly non-linear, the Sequential
objective and design variables are established. Chatter Quadratic Programming (SQP) method is used in

Fig. 9. Implication process for rule 4.

 O. Maeda et al. / International Journal of Machine Tools & Manufacture 45 (2005) 537–548 543

Fig. 10. Fuzzy logic process for transmission selection (gear/belt-pulley).

the optimization of the spindle design. The iterative defines the distance moving in direction d. The optimization
optimization operations can be expressed as the following algorithm used in this paper is shown in Fig. 11.
equation Objective function. The cutting conditions, the depth of
cut and spindle speed, must be under the stability lobes in
xkC1 Z xk C x* d (8) order to avoid chatter vibrations in metal cutting [16].
The location of stability pockets, the lobes, is dependent on
the natural frequencies of the spindle system, and the
where k is the iteration number, xkC1 is the new design allowable depth of cut depends on the dynamic stiffness of
variable vector, xk is the current design variable vector, d is a the modes. This paper proposes automated tuning of the
vector search direction, and a* is the scalar quantity that spindle modes in such a way that chatter vibration free
pockets of stability is created at the desired spindle speed
and depth of cut. The tuning of the spindle dynamics is
achieved by optimizing the distribution of bearings along
the spindle shaft.
When the critical depth of cut of the stability lobes at the
cutting spindle speed (aClim) (Fig. 12) is maximized,

Fig. 11. Optimization algorithm. Fig. 12. Stability lobe of milling.

544 O. Maeda et al. / International Journal of Machine Tools & Manufacture 45 (2005) 537–548

Fig. 13. Design variables for bearing locations.

the cutting is the most stable. Therefore, the objective Eq. (10) is decoupled as an SDOF system
function (fob) is simply defined as follows
mi y€ i C ci y_i C ki yi Z fi ; i Z 1; .; m (12)
X
N
Minimize : fob Z K Wi ðaClim Þi ; i Z 1; 2; .; Nf (9) 
 C;
where mi, ci, ki are the diagonal elements of matrices M;
iZ1

K; respectively.
where Wi and (aClim)i are the weight and critical depth of cut The transfer function for Eq. (12) is
for the ith cutter, respectively, and Nf is the total number of
cutters with different flutes. y ðsÞ 1=mi
Gi ðsÞ Z i Z 2 ; i Z 1; .; m (13)
In order to calculate aClim, the Frequency Response f i ðsÞ s C 2zi un;i s C un;i
Function (FRF) at the tool tip is required, which is evaluated
by Timoshenko Beam based Finite Element model of the where p Z ciffi=ð2mi un;i Þ is the damping ratio for the ith mode;
ziffiffiffiffiffiffiffiffiffi
spindle. un;i Z ki =mi is the ith natural frequency.
Design variables. The design variables for the chatter From Eq. (13)
free spindle design are the bearing spans. The number of the
design variables depends on the bearing arrangement and yi ðsÞ Z Gi ðsÞfi ðsÞ; i Z 1; .; m (14)
the transmission type. Fig. 13 shows the design variables of
the motorized spindle with five bearings (four front and one In matrix form, Eq. (14) becomes
rear bearing). The x(1) to x(6) are the design variables which

YðsÞ Z GðsÞFðsÞ (15)
define the bearing locations. The x is determined automati-
cally with the subtraction of the other known length from 
Due to XðsÞZ PYðsÞZ PGðsÞFðsÞZ PGðsÞPT FðsÞ; the
the total spindle length.
transfer function for the spindle system is obtained as
Initial conditions. In order to prevent optimization to
follows:
converge to one local minimum as opposed to a global
minimum, three different initial conditions are used and the XðsÞ
final design is chosen from the three emerging optimal HðsÞ Z Z PGðsÞPT (16)
FðsÞ
solutions.
Calculation of FRF at the tool tip. In order to compute The element in the transfer function matrix H(s) is
the FRF at the tool tip, which is required for the chatter
stability analysis, the tool holder and cutter are taken into xj ðsÞ Xm Xm
uji uki
hjk Z Z Pji Pki Gi ðsÞ Z
account in the specifically developed FE model where the Fk ðsÞ s C 2zi un;i s C u2n;i
2
iZ1 iZ1
damping ratio of each mode is set to 0.03, which is typical in
(17)
the mechanical system. The system uses an in house
developed FE algorithm based on standard Timoshenko where uji is the element of the mass normalized mode
Beam model. shape u.
By using the FE method the following forced vibration Calculation of critical depth of cut for chatter stability
equation can be obtained for the spindle system: (aClim): the milling chatter stability theory developed by
Mx€ C Cx_ C Kx Z F (10) Budak–Altintas [16] is used to evaluate critical depth
  2 
In modal coordinates, it can be expressed as, K2pLR LI
alim Z 1C (18)
 C Ky
 y€ C Cy
M  Z F (11) NKtc LR

where MZ  PT MP; CZ  PT CP and KZ  PT KP are the where Ktc is the cutting coefficient, and N is the number of
modal mass matrix, damping matrix, and stiffness matrix, teeth. LR and LI are real and imaginary values of an
respectively. These are all m!m diagonal matrices; FZ eigenvalue L, which is obtained using the FRF at the tool
PT F is modal force; xZPy, where P is mode shape which is tip. The details of the chatter stability theory can be found
an n!m matrix. in [11,16].
 O. Maeda et al. / International Journal of Machine Tools & Manufacture 45 (2005) 537–548 545

transmission with oil–air type lubrication with four bearings

at the front and one at the rear. The maximum spindle speed
is 20,000 rpm and the power and torque properties of the
spindle motor are set from the data shown in [17]. It is
assumed that the user wishes to use the machine
predominantly in cutting aircraft parts made from
Al7075-T6 with a four-fluted end mill with a desired
depth of cut of 3 mm and 15,000 rpm spindle speed. The
most common cutting conditions are listed in Table 4.
From the power and torque diagram of the spindle motor
of SH-403, the motor torque (Tmax c) at the cutting
Fig. 14. Spindle configuration of SH-403.
speed 15,000 rpm is found to be as 8.8 N m. On the other
hand, the cutting torque (Tmo) required for the cutting
Table 4 conditions shown in Table 4 can be calculated as 4.796 N m.
Most common cutting conditions for SH-403
The maximum motor rotation speed (nmax) is 20,000 rpm
Cutter diameter 19.05 mm and the desired speed (n) is 15,000 rpm. Therefore,
Number of flutes 4 Tmax c%Tmo and n%nacmax(Znmax), so either the direct
Material to be cut AL7075-T6
Cutting spindle speed 9000 rpm coupling type or the motorized type are the possible
Depth of cut 3 mm transmission type.
Width of cut 19.05 mm (slotting) In order to select the transmission type from either the
Feed rate 0.1 mm/flute direct coupling type or the motorized type, ‘Spindle Speed’,
‘High Dynamic Stiffness vs. Low Balancing Vibration’,
4. Application of the expert spindle design system ‘Low Thermal Effect vs. Small Noise’, and ‘Low Replace-
ment Operation Cost vs. Low Replacement Parts Cost’ need
The proposed system is demonstrated against a commer- to be input. The spindle speed is automatically set from the
cially existing machine tool (Mori Seiki SH-403) as shown maximum motor speed. In this case, the numbers are
in Fig. 14 for comparison. The main spindle specifications assumed from the concepts of the machine written in the
of SH-403 are shown in [17]. The spindle has a motorized SH-403 catalog [17]. The fuzzy weight numbers are set

Table 5
Selection Results for Mori-Seiki SH-403

Table 6
The result of the spindle components’ selection

Case No. Spindle name Actual design Expert system

Case 1 Mori Seiki SH-403 [17] T Motorized Motorized
L Oil air Oil air
Case 2 Starragheckert 260-50-24 [18] T Motorized Motorized
L Oil air Oil air
Case 3 Weiss Spindle MAL UBC Version T Belt-pulley Belt-pulley
L Grease Grease
Case 4 Mitsui Seiki VU65A [19] T Gear Gear
L Oil Air Oil Air
Case 5 Matsuura H. 260-50-24 [20] T D. coupling D. coupling
L Grease Grease

T, transmission; L, lubrication; D. coupling, direct coupling

546 O. Maeda et al. / International Journal of Machine Tools & Manufacture 45 (2005) 537–548

Fig. 15. Design constants for optimization of bearing locations.

Fig. 16. Design constraints.

as ‘3’, ‘8’, and ‘4’, respectively. Similarly, the fuzzy weight Table 6 shows the list of agreements between the
of the lubrication system is set. actual design and the design that uses this expert system.
Table 5 shows the results of the spindle components’ The results attained via the expert system match those of the
selection. The Expert Spindle Design System selected the actual design in all five cases. Therefore, the proper spindle
proper transmission and lubrication type. components can be selected with the proposed expert

Fig. 17. Initial conditions of bearing locations.

 O. Maeda et al. / International Journal of Machine Tools & Manufacture 45 (2005) 537–548 547

The SQP is used to optimize the bearing spans. The

gradients objective function fob and Hessian matrix cannot
be calculated analytically. They are therefore obtained using
a numerical differentiation and Quasi-Newton methods,
respectively. The minimum change in the design variables
for finite difference gradients is set to 0.01 mm so that the
first partial derivative Vfob is derived by subtracting
0.01 mm from the bearing spans. The BFGS approximation
technique is used to update the Hessian matrix. The
termination tolerance of the objective function values is
set to 0.0001. In the case that the optimized value difference
between the previous and current values is less than 0.0001,
the calculation stops computing and the final value is shown
as the optimized value.
Fig. 18 shows the chatter stability lobes for a four flute
Fig. 18. Stability lobes for initial and optimized designs with single cutter.
cutter computed from the three initial designs shown in
system, that is, the rule base and the membership functions Fig. 17 and the final optimized design. The desired spindle
of this system are defined properly. speed is 9000 rpm, and the depth of cut is 3 mm. The cutting
is not stable for all three initial designs, but it becomes
4.1. Optimization for bearing locations stable after optimization. The optimized spindle configur-
ation is shown in Fig. 19.
The bearing locations are optimized with the same cutting Multiple cutters with different flutes can also be optimized
conditions shown in Table 4. The design constants and for the same spindle. The same spindle and cutting conditions
bearing parameters are set suitably as shown in Fig. 15, described above are used, but three cutters with two, three,
which are not the dimensions of the Mori Seiki SH-403. and five flutes are used on the spindle. Only the first initial
The bearing spans’ constraints are set by considering the design is considered here. Fig. 20a and b show the stability
spaces of lubrication devices as shown in Fig. 16. lobes for the original and optimized designs, respectively. In
Three types of initial bearing spans are set without the original design, the cutting is not stable for the five-fluted
violating the constraints. The following three spans are used cutter, and it is close to the unstable regions for two and three
as the initial spans, as shown in Fig. 17. fluted cutters. For the optimized design, there are very big

Fig. 19. Optimized spindle model with single cutter.

Fig. 20. (a) Stability lobes for the initial design with multiple cutters. (b) Stability lobes for the optimized design with multiple cutters.
548 O. Maeda et al. / International Journal of Machine Tools & Manufacture 45 (2005) 537–548

Fig. 21. Optimized spindle model with multiple cutters.

margins for two and three fluted cutters, and the cutting design strategy is demonstrated in designing several
becomes stable for the five-fluted cutter although it is close to industrial size spindles used in industry.
the unstable region. The optimized spindle configuration is
shown in Fig. 21.
References

[1] H. Eskicioglu, M. Akkok, O.H. Yildiz, Computer aided selection of machine

5. Conclusions
tool spindle bearing arrangements, American Society of Mechanical
Engineers, Petroleum Division (Publication) PD 64 (8-1) (1994) 181–187.
This paper presents an Expert Spindle Design system for [2] W.L.D. Wong, J. Atkinson, A knowledge cell approach to processing
machine tool engineers. It proposes an alternative method to design information, Journal of Materials Processing Technology 107
the present design practice, which is based on the past (1–3) (2000) 44–52.
[3] S. Yang, A study of the static stiffness of machine tool spindles,
experience of individual designers, while attempting to International Journal of Machine Tool Design & Research 21 (1)
eliminate costly trials by using the laws of machine design, (1981) 23–40.
solid mechanics, and metal cutting dynamics in an [4] S. Taylor, B.T. Khoo, D. Walton, Microcomputer optimization of
integrated fashion. machine tool spindle stiffnesses, International Journal of Machine
Tools and Manufacture 30 (1) (1990) 151–159.
The design configurations and membership functions are
[5] D. Lee, D. Choi, Reduced weight design of a flexible rotor with ball
stored in a knowledge base using sets of design rules based bearing stiffness characteristics varying with rotational speed and
on the past experience and laws of cutting mechanics. Fuzzy load, Journal of Vibration and Acoustics, Transactions of the ASME
logic is used as an inference engine in the proposed expert 122 (3) (2000) 203–208.
system. The fuzzy logic can deal with design uncertainties [6] W.J. Chen, M. Rajan, S.D. Rajan, H.D. Nelson, Optimal design of
squeeze film dampers for flexible rotor systems, Journal of
such as high, medium and low speeds or large/small torque
Mechanisms, Transmissions, and Automation in Design 110 (2)
required from the spindles where the exact numerical values (1988) 166–174.
are difficult to set rigidly by the designers. The membership [7] C. Nataraj, H. Ashrafiuon, Optimal design of centered squeeze film
functions can be updated by the designers as the rules dampers, Journal of Vibration, Acoustics, Stress, and Reliability in
change due to technological advances in industry. The Design 115 (2) (1993) 210–214.
[8] W.R. Wang, C.N. Chang, Dynamic analysis and design of a machine
expert system leads to automatic generation of spindle tool spindle-bearing system, Journal of Vibration and Acoustics,
configuration which includes drive shaft, motor type and Transactions of the ASME 116 (3) (1994) 280–285.
size, transmission mechanism between the motor and shaft, [9] G. Hagiu, M.D. Gafiranu, Preload optimization: a high efficiency
and tool holder style. design solution for grinding machines main spindles, ASTM Special
While the configuration and sub-components of the Technical Publication, No. 1247, 1995.
[10] Y. Kang, Y.P. Chang, J.W. Tsai, S.C. Chen, L.K. Yang,
spindle are based on the torque, power, and speed C.A.E. Integrated, Strategies for the design of machine tool spindle-
requirements from the machine tool, the exact locations of bearing systems, Finite Elements in Analysis and Design 37 (6/7)
the bearings must be determined based on the chatter (2001) 485–511.
vibration stability of the spindle. The paper proposes a [11] Y. Altintas, Manufacturing automation-metal cutting mechanics,
Machine Tool Vibrations, and CNC Design, Cambridge University
bearing spacing optimization strategy for the spindles
Press, Cambridge, 2000.
configured by the expert system or designed by the [12] T. Moriwaki, et al., Kousakukikai no sekkeigaku (kiso-hen), The
engineers. The designer provides initial estimates of the Design for Machine tools (Basics), Japan Machine Tool Builder’s
bearing locations including constraints. The configured Association, 1998.
spindle is analyzed by a proposed Finite Element Analysis [13] FAG Ultra High Speed and Super Precision Spindle Bearings,
Fukuda Corp.
(FEA) algorithm based on Timoshenko Beam elements. The [14] NSK Ltd, Precision Rolling Bearings for Machine Tool Spindles,
Frequency Response Function (FRF) of the spindle at the 1999.
tool tip is obtained by the modal analysis module of the FEA [15] NSK-RHP, RHP Super Precision Bearings, 1998.
algorithm. The bearing locations are optimized iteratively [16] Y. Altintas, E. Budak, Analytical Prediction of Stability Lobes in Milling,
until the designed spindle satisfies chatter vibration free CIRP Annals—Manufacturing Technology 44 (1) (1995) 357–362.
[17] SH-403, MORI SEIKI CO., LTD.
cutting constraints. Sequential Quadratic Programming [18] High Performance Motor Spindles, Starragheckert.
(SQP) is used as an optimization method in identifying [19] VU65A Vertical Machining Center, Mitsui Seiki Kogyo Co., Ltd.
optimal bearing locations. The proposed expert spindle [20] H. Max Series, Matsuura Machinery Co.